PreprintPDF Available

Climate change will drive novel cross-species viral transmission

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Between 10,000 and 600,000 species of mammal virus are estimated to have the potential to spread in human populations, but the vast majority are currently circulating in wildlife, largely undescribed and undetected by disease outbreak surveillance [1,2,3]. In addition, changing climate and land use drive geographic range shifts in wildlife, producing novel species assemblages and opportunities for viral sharing between previously isolated species [4,5]. In some cases, this will inevitably facilitate spillover into humans [6,7] - a possible mechanistic link between global environmental change and emerging zoonotic disease [8]. Here, we map potential hotspots of viral sharing, using a phylogeographic model of the mammal-virus network, and projections of geographic range shifts for 3,870 mammal species under climate change and land use scenarios for the year 2070. Shifting mammal species are predicted to aggregate at high elevations, in biodiversity hotspots, and in areas of high human population density in Asia and Africa, sharing novel viruses between 3,000 and 13,000 times. Counter to expectations, holding warming under 2 C within the century does not reduce new viral sharing, due to greater range expansions - highlighting the need to invest in surveillance even in a low-warming future. Most projected viral sharing is driven by diverse hyperreservoirs (rodents and bats) and large-bodied predators (carnivores). Because of their unique dispersal capacity, bats account for the majority of novel viral sharing, and are likely to share viruses along evolutionary pathways that could facilitate future emergence in humans. Our findings highlight the urgent need to pair viral surveillance and discovery efforts with biodiversity surveys tracking range shifts, especially in tropical countries that harbor the most emerging zoonoses.
Content may be subject to copyright.
Climate change will drive novel cross-species1
viral transmission2
Colin J. Carlson1,2,, Gregory F. Albery1,3,, Cory Merow4, Christopher H.3
Trisos5, Casey M. Zipfel1, Evan A. Eskew3, Kevin J. Olival3, Noam Ross3, and4
Shweta Bansal1
5
1Department of Biology, Georgetown University, Washington, D.C., USA.6
2Center for Global Health Science & Security, Georgetown University, Washington, D.C., USA.7
3EcoHealth Alliance, New York, NY, USA.8
4Ecology and Evolutionary Biology, University of Connecticut, Storrs, CT, USA.9
5African Climate and Development Initiative, University of Cape Town, Cape Town, South10
Africa.11
These authors share equal authorship.12
January 24, 202013
1
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Abstract14
Between 10,000 and 600,000 species of mammal virus are estimated to have the15
potential to spread in human populations, but the vast majority are currently cir-16
culating in wildlife, largely undescribed and undetected by disease outbreak surveil-17
lance1,2,3. In addition, changing climate and land use drive geographic range shifts18
in wildlife, producing novel species assemblages and opportunities for viral sharing19
between previously isolated species4,5. In some cases, this will inevitably facilitate20
spillover into humans6,7—a possible mechanistic link between global environmental21
change and emerging zoonotic disease8. Here, we map potential hotspots of viral22
sharing, using a phylogeographic model of the mammal-virus network, and projec-23
tions of geographic range shifts for 3,870 mammal species under climate change and24
land use scenarios for the year 2070. Shifting mammal species are predicted to ag-25
gregate at high elevations, in biodiversity hotspots, and in areas of high human pop-26
ulation density in Asia and Africa, sharing novel viruses between 3,000 and 13,00027
times. Counter to expectations, holding warming under 2°C within the century28
does not reduce new viral sharing, due to greater range expansions—highlighting29
the need to invest in surveillance even in a low-warming future. Most projected vi-30
ral sharing is driven by diverse hyperreservoirs (rodents and bats) and large-bodied31
predators (carnivores). Because of their unique dispersal capacity, bats account for32
the majority of novel viral sharing, and are likely to share viruses along evolutionary33
pathways that could facilitate future emergence in humans. Our findings highlight34
the urgent need to pair viral surveillance and discovery efforts with biodiversity35
surveys tracking range shifts, especially in tropical countries that harbor the most36
emerging zoonoses.37
2
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Main Text38
In the face of rapid environmental change, survival for many species depends on moving39
to track shifting climates. Even in a best case scenario, many species are projected40
to shift a hundred kilometers or more in the next century9,10. In the process, many41
animals will bring their parasites and pathogens into new environments4,11, creating new42
evolutionary opportunities for host jumps8. Most conceptual frameworks for cross-species43
transmission revolve around how these host jumps facilitate the spillover of new zoonotic44
pathogens into humans12,13,14 , but viral evolution is an undirected process15, in which45
humans are only one of over 5,000 mammal species with over 12 million possible pairwise46
combinations16 . Despite their indisputable significance, zoonotic emergence events are47
just the tip of the iceberg; almost all cross-species transmission events will occur among48
wild mammals, largely undetected and mostly inconsequential for public health.49
Of the millions of possible pairwise viral exchanges, the vast majority are biologically50
implausible, as host species’ geographic ranges currently do not overlap. However, as51
ranges shift, a small fraction of possible interactions will occur, of which a subset will52
lead to viral establishment in a novel host. Which subset results in establishment de-53
pends on opportunity and compatibility 14,17,18, analogous to exposure and susceptibility54
within populations, and both dimensions pose an important predictive challenge. The55
ability of species to track shifting habitats in a changing climate will determine which56
pairs of species encounter each other for the first time4,19 . Habitat selection and be-57
havioral differences can further limit contact, even if species are nominally sympatric19.58
Some viruses may spread environmentally between spatially-proximate species with no59
direct behavioral contact20 , but generally, sharing is more likely among species with more60
ecological overlap21 . Even among species in close contact, most spillovers are still a dead61
end; progressively smaller subsets of viruses can infect novel host cells, proliferate, and62
transmit onward in a new host18 . Their ability to do so is determined by compatibility63
between viral structures, host cell receptors, and host immunity6. Because closely related64
species share both ecological and immunological traits through identity by descent, phy-65
logeny is a strong predictor of pathogen sharing17,22 , as well as susceptibility to invasion66
by new viruses23,24,25 . In a changing world, these factors should continue to mediate the67
impact of ecosystem turnover on the mammalian virome.68
Although several studies have mapped current hotspots of emerging diseases3,26,27,69
few have modeled them in the context of global change. With the global reassortment70
of animal biodiversity, it is unknown whether bats and rodents will still play a central71
role in viral emergence3,28 (ED Figure 1), or whether hotspots of viral emergence will72
stay in tropical rainforests27,29 which currently harbor most undiscovered viruses3,30.73
Here, by projecting geographic range shifts and applying fundamental biological rules74
3
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
for cross-species transmission, we predicted how and where global change could create75
novel opportunities for viral sharing. We built species distribution models for 3,87076
mammal species, and projected geographic range shifts based on four paired scenarios77
of climate change (representative concentration pathways, RCPs) and land use change78
(shared socioeconomic pathways, SSPs) by 2070. We treated dispersal potential as an79
additional layer of biological realism, inferring these limits for species based on allomet-80
ric scaling31, and compared predictions with and without dispersal constraints. We used81
these projections to identify where novel range overlap among unfamiliar species (“first82
encounters”) could happen, and used a recently-developed model to predict the proba-83
bility of viral sharing based on geographic overlap and host phylogenetic similarity17.84
(ED Figure 2) This model framework allows powerful inference based on the 1% of85
the global mammalian virome that has been described 1,3,17 . Using this approach, we86
tested the hypothesis that environmental change should drive biotic homogenization of87
mammal communities, exposing mammals to novel viruses, and altering the structure of88
mammal-virus interactions.89
Most mammals are projected to undergo rapid range shifts in the next half century10.90
If range shifts can keep pace with the velocity of climate change32, we predict that the91
vast majority of mammal species (89%–98%) will overlap with at least one unfamiliar92
species somewhere in their future range, regardless of emissions scenario. At the global93
level, community turnover would permit almost 300,000 novel species interactions (ED94
Figure 3). These “first encounters” between mammal species will occur everywhere in the95
world, but are concentrated in tropical Africa and southeast Asia (ED Figure 4). This96
result was surprising, and counter to our expectation that species might aggregate at97
higher latitudes, given that most research has focused on poleward range shifts33,34,35,98
and previous work has anticipated a link between climate change, range shifts, and99
parasite host-switching in the Arctic36,37 . However, our findings show that communities100
tend to shift along latitudinal gradients together, with species rarely encountering new101
conspecifics38. In contrast, species will track thermal optima along elevational gradents102
and aggregate in novel combinations in mountain ranges, especially in tropical areas with103
the highest baseline diversity, matching prior predictions39.104
This global re-organization of mammal assemblages is projected to dramatically im-105
pact the structure of the mammalian virome. Accounting for geographic opportunity106
and phylogenetic compatibility, we projected that a total of 279,427 first encounters in107
RCP 2.6 would lead to nearly 12,000 novel sharing events. Assuming that spillover will108
be localized to areas of novel host overlap, we mapped expected viral sharing events, and109
found again that most sharing should occur in high-elevation, species-rich ecosystems110
in Africa and Asia (Figure 1A). If species survive a changing climate by aggregating in111
high elevation refugia, this suggests emerging viruses may be an increasing problem for112
4
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
their conservation40,41 . Across scenarios, the spatial signal of expected sharing events is113
nearly identical, and dominated more by the extent of range shifts than by underlying114
community phylogenetic structure (ED Figure 5); at least in our framework, opportunity115
drives spatial patterns more than compatibility.116
Species’ dispersal capacity is likely to constrain range shifts, and therefore to limit117
novel viral exchange. We limited the dispersal potential of flightless species further to the118
restrictions placed on the SDM projections, based on an established allometric scaling119
with body size, trophic rank, and generation time 42 . Dispersal limits caused significant120
reductions in range expansions across all scenarios, especially warmer ones, and therefore121
drove a reduction in novel interactions. Even in RCP 2.6 (the mildest scenario), limiting122
dispersal reduced the number of first encounters by 60%, and reduced the associated viral123
sharing events by 69%—to a still-staggering 3,600–3,800 projected viral sharing events.124
Because trophic position and body size determine dispersal capacity, carnivores account125
for a disproportionate number of first encounters, while ungulates and rodents have126
slightly fewer first encounters than expected at random (ED Figure 6) Spatial patterns127
also changed dramatically when dispersal constraints were added, with the majority of128
first encounters and cross-species viral transmission events occurring in southeast Asia129
(Figure 1B, ED Figures 4, 5). This viral sharing hotspot is driven disproportionately130
by bats, because their dispersal was left unconstrained; we made this choice given their131
exclusion from the original study31, genetic evidence that flight allows bats—and their132
viruses—to circulate at continental levels 43,44 , and data suggesting that bat distributions133
are already undergoing disproportionately rapid shifts45. Bats account for 87% of first134
encounters after constraining dispersal, and dominate the spatial pattern, with most of135
their first encounters restricted to southeast Asia (Figure 2).136
Bats’ unique capacity for flight could be an important and previously unconsidered137
link between climate-driven range shifts and future changes in the mammal virome.138
Even non-migratory bats can regularly travel hundreds of kilometers within a lifetime,139
far exceeding what small mammals might be able to cover in 50 years; half of all bat140
population genetic studies have failed to find any evidence for isolation by distance46.141
This unique dispersal capacity has inevitable epidemiological implications, with recent142
evidence suggesting that continental panmixia may be common for zoonotic reservoirs,143
and allow viral circulation at comparable scales43,44,47 . We found that a staggering144
number of studies have also identified ongoing rapid range expansions in bat species145
around the world45,48,49,50,51,52,53,54,55 , with little mention in the broader climate change146
or emerging disease literature. If flight does allow bats to undergo more rapid range147
shifts than other mammals, we expect they should drive the majority of novel cross-148
species viral transmission, and likely bring zoonotic viruses into new regions. This could149
add an important new dimension to ongoing debate about whether bats are “special”150
5
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
due to their higher viral richness, higher proportion of zoonotic viruses, and potentially151
unique immune adaptations3,56,57,58,59 .152
More broadly, climate-driven changes in the mammalian virome are likely to cascade153
in future emergence of zoonotic viruses. Among the tens of thousands of expected viral154
host jumps, some of the highest-risk zoonoses or potential zoonoses are likely to find155
new hosts. This may pose a threat to human health down the road: the same general156
rules for cross-species transmission explain spillover patterns for emerging zoonoses60,61,157
and the viral species that make successful jumps across wildlife species have the highest158
propensity for zoonotic emergence3,7,28. Just as simian immunodeficiency virus emer-159
gence in chimpanzees and gorillas facilitated the origin of HIV, or SARS-CoV spillover160
into civets allowed a bat virus to reach humans, these wildlife-to-wildlife host jumps may161
be evolutionary stepping stones for the 10,000 to 600,000 potentially zoonotic viruses162
that are currently circulating in mammal hosts1.163
To illustrate this problem, we constructed a sub-network of 13 possible Zaire ebolavirus164
hosts in Africa, and projected possible first encounters involving these species (Figure165
3A-C). We project these 13 species to encounter 3,604 new mammals in RCP 2.6, with166
a modest reduction to 2,586 species by dispersal limits. These first encounters are pre-167
dicted to produce 87 new viral sharing events that might include ZEBOV, and which168
cover a much broader part of Africa than the current zoonotic niche of Ebola 62 . Hu-169
man spillover risk aside, this could expose several new wildlife species to a deadly virus,170
historically responsible for sizable primate die-offs63. Moreover, for zoonoses like Zaire171
ebolavirus without known reservoirs, future host jumps would only complicate urgent172
efforts to trace the source of spillover and anticipate future emergences 64,65 . Ebola is173
far from unique: with 5,762–11,122 first encounters between bats and primates alone174
leading to an expected 57–181 new viral sharing events across scenarios (Figure 3D),175
many potential zoonoses are likely to experience new evolutionary opportunities because176
of climate change.177
Future hotspots of novel assemblages and viral evolution are projected to coincide178
areas of high human population density, further increasing vulnerability to potential179
zoonoses. First encounters are disproportionately likely to occur in areas that are pro-180
jected to be either human settled or used as cropland, and surprisingly less likely to181
occur in forests, which current literature highlights as producing most emerging diseases182
(Figure 4)27. This finding is consistent for bats and non-bats, and may be an accident183
of geography, but more likely represents the tendency of human settlements to aggre-184
gate on continental edges and around biodiversity hotspots66. Regardless of mechanism,185
we predict that tropical hotspots of novel viral sharing will broadly coincide with high186
population density areas in 2070, especially in the Sahel, the Ethiopian highlands and187
the Rift Valley, India, eastern China, Indonesia, and the Philippines (Figure 4). Some188
6
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
European population centers also land in these hotspots; recent emergences in this re-189
gion like Usutu virus67 highlight that these populations can still be vulnerable, despite190
greater surveillance and healthcare access. If range-shifting mammals create ecological191
release for undiscovered zoonoses, populations in these areas are likely to be the most192
vulnerable.193
Whereas most studies agree that climate change mitigation through reducing green-194
house gas emissions will prevent extinctions and minimize harmful ecosystem impacts,195
our results suggest that mitigation cannot reduce the likelihood of climate-driven viral196
sharing. Instead, the mildest, slowest scenarios for biotic homogenization appear likely197
to produce the most cross-species viral transmission: when climate velocity is lowest,198
species can successfully track shifting climate optima, leading to more range expansion,199
and more first encounters. Accounting for dispersal limits, species gained an average200
of 75% range in the mildest pathway (RCP 2.6); in comparison, only 28% of species201
experienced a net expansion in the most extreme pathway (RCP 8.5), for an average of202
21% range gain. (ED Figure 3A) In fact, in the warmest scenario, up to 326 species lost203
their entire range, with 168 attributable to dispersal limits alone. As a result, there were204
5% fewer first encounters in RCP 8.5 compared to RCP 2.6, and unexpectedly, a 2%205
reduction in the connectivity of the future global sharing network. (ED Figure 3B,D)206
Overall, our results indicate that a mild perturbation of the climate system could create207
thousands of new eco-evolutionary opportunities for viruses. We caution that this does208
not imply a possible upside to catastrophic warming, which will be accompanied by mass209
defaunation, devastating disease emergence, and unprecedented levels of human displace-210
ment and global instability. Rather, our results highlight the urgency of better wildlife211
surveillance systems and health infrastructure as a form of climate change adaptation,212
even if mitigation efforts are successful and global temperatures stay under +2°C.213
Our study establishes a macroecological link between climate change and cross-species214
viral transmission. In practice, the patterns we describe are likely to be complicated by215
several ecological factors, including the temperature sensitivity of viral host jumps68;216
the possibility that defaunation especially at low elevations might interact with disease217
prevalence through biodiversity dilution and amplification effects, not captured by our218
models69; or temporal heterogeneity in exposure (hosts might exchange viruses in passing219
but not overlap by 2070, especially in warmer scenarios). Future work can also expand220
the scope of our findings to other host-parasite systems; our novel approach, which221
combines viral sharing models with massive species distribution modeling pipelines, is222
readily applied to other datasets. Birds have the best documented virome after mammals,223
and changing migration targets in a warming world may be especially important targets224
for prediction. With amphibians facing disproportionately high extinction rates due225
to a global fungal panzootic, and emerging threats like ranavirus causing conservation226
7
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
concern, viral exchange among amphibians may be especially important information for227
conservation practitioners70 . Finally, marine mammals are an important target given228
their exclusion here, especially after a recent study implicating reduced Arctic sea ice in229
viral sharing among sympatric pinnipeds and sea otters—a result that may be the first230
proof of concept for our proposed climate-disease link 71 .231
Because hotspots of cross-species transmission are predictable, our study provides232
the first template for how surveillance could target future hotspots of viral emergence in233
wildlife. In the next decade alone, over a billion dollars could be spent on a proposed234
global effort to identify zoonotic threats before they spread from wildlife reservoirs into235
human populations2. These efforts are being undertaken during the greatest period236
of global ecological change recorded in human history, and in a practical sense, the237
rapid movement of species and formation of no-analog communities poses an unexpected238
challenge for virological research. While several studies have addressed how range shifts239
in zoonotic reservoirs might expose humans to novel viruses, few have considered the fact240
that most new exposures will be among mammal species. Tracking spillover into humans241
is paramount, but so is tracking of viral sharing in wildlife, and targeting surveillance in242
hotspots of future sharing may help researchers identify host jumps early on.243
8
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Methods244
In this study, we develop global maps for terrestrial mammals that model their eco-245
logical niche as a function of climate and habitat use. We project these into paired246
climate-land use futures for 2070, with dispersal limitations set by biological constraints247
for each species. We predict the probability of viral sharing among species pairs us-248
ing a model of the mammalian viral sharing network that is trained on phylogenetic249
relatedness and current geographic range overlaps. With that model, we map the pro-250
jected hotspots of new viral sharing in different futures. All analysis code is available at251
github.com/cjcarlson/iceberg.252
Mapping species distributions253
We developed species distribution models for a total of 3,870 species in this study, divided254
into two modeling pipelines based on data availability (ED Figures 8, 9).255
Data Collection256
We scraped the Global Biodiversity Informatics Facility (GBIF) for mammal occurrence257
records, and developed species distribution models for all 3,870 species with at least 3258
unique terrestrial presence records on a 25 km by 25 km grid (one unique point per grid259
cell). This grain was chosen based on the availability of future land use projections (see260
below). Spatial and environmental outliers were removed based on Grubb outlier tests261
(p-value of 1e-3)72 .262
Poisson point process models263
For 3,088 species with at least 10 unique presence records, Poisson point process models264
(closely related to Maxent) were fit using regularized downweighted Poisson regression73
265
with 20,000 background points fit with the R package glmnet 74,75,74 . The spatial do-266
main of predictions was chosen based on the continent(s) where a species occurred in267
their IUCN range map. We trained species distribution models on current climate data268
using the WorldClim 2 data set 76 , using mean annual temperature, mean diurnal temper-269
ature range, annual precipitation, precipitation seasonality, and precipitation in warmest270
quarter/ (precipitation in warmest quarter + precipitation in coldest quarter). These271
predictors were chosen based on having global correlations <0.7 among one another.272
These candidate predictors were further filtered on a species-by-species basis, retaining273
the maximum number of predictors with correlation <0.7 within the domain where the274
model was fit.275
9
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Models were fit with 5-fold cross validation, where folds were assigned based on spa-276
tial clusters to remove the influence of spatial autocorrelation on cross-validation perfor-277
mance statistics. Linear (all species), quadratic (species with >100 records), and product278
(species with >200 records) features were used. The regularization parameter was de-279
termined based on 5-fold cross-validation with each fold, choosing a value 1 standard280
deviation below the minimum deviance77 . This resulted in five models per species which281
were then combined in an unweighted ensemble. Continuous predictions of the ensemble282
were converted to binary presence/absence predictions by choosing a threshold based on283
the 5th percentile of the ensemble predictions at training presence locations.284
When models were projected into the future, we limited extrapolation to 1 standard285
deviation beyond the data range of presence locations for each predictor. This decision286
balances a small amount of extrapolation based on patterns in a species niche with287
limiting the influence of monotonically increasing marginal responses, which can lead to288
statistically unsupported (and likely biologically unrealistic) responses to climate.289
Range bagging models290
For an additional 783 rare species (3 to 9 unique points on the 25 km grid), we produced291
species distribution models with a simpler range bagging algorithm, a stochastic hull-292
based method that can estimate climate niches from an ensemble of underfit models78,79,293
and is therefore well suited for smaller datasets. From the full collection of presence294
observations and environmental variables range-bagging proceeds by randomly sampling295
a subset of presences (proportion p) and a subset of environmental variables (d). From296
these, a convex hull around the subset of points is generated in environmental space. The297
hull is then projected onto the landscape with a location considered part of the species298
range if its environmental conditions fall within the estimate hull. The subsampling is299
replicated Ntimes, generating N‘votes’ for each cell on the landscape. One can then300
choose a threshold for the number of votes required to consider the cell as part of the301
species’ range to generate the binary map used in our downstream analyses. Based on302
general guidelines in78 we chose p= 0.33,d= 2, and N= 100. We then chose the voting303
threshold to be 0.165 (=0.33/2) because this implies that the cell is part of the range304
at least half the time for each subsample. Upon visual inspection, this generally lead to305
predictions that were very conservative about inferring that unsampled locations were306
part of a species distribution. The same environmental predictors and ecoregion-based307
domain selection rules were used for range bagging models as were used for the point308
process models discussed above. This hull-based approach is particularly valuable for309
poorly sampled species which may suffer from sampling bias because bias within niche310
limits has little effect on range estimates.311
10
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Model validation312
PPM models performed well, with a mean test AUC under 5 fold cross-validation (using313
spatial clustering to reduce inflation) of 0.77 (s.d. 0.13). The mean partial AUC eval-314
uated over a range of sensitivity relevant for SDM (0.8-0.95) was 0.8 (s.d. 0.08). The315
mean sensitivity of binary maps used to assess range overlap (based on the 5% training316
threshold used to make a binary map) was 0.89 (s.d. 0.08). Range bagging models were317
difficult to meaningfully evaluate because they were based on extremely small sample318
sizes (3-9). The mean training AUC (we did not perform cross-validation due to small319
sample size) was 0.96 (s.d. 0.09). The binary maps had perfect sensitivity (1) because320
the threshold used to make them was chosen sufficiently low to include the handful of321
known presences for each species. One way to assess how much we inferred the range322
for these species is to quantify how much of the range was estimated based on out mod-323
els, based on the number of (10km) cells predicted to be part of the species range even324
when it was not observed there. The mean number of cells inferred to contain a presence325
was 253 (s.d. 448); however, the distribution is highly right skewed with a median of326
94. This indicates that the range bagging models were typically relatively conservative327
about inferring ranges for poorly sampled species.328
Habitat range and land use329
We used the Land Use Harmonization version 2.0 (LUH2) gridded dataset to capture330
global patterns in land cover for the present and future 80 . These data are derived from331
an integrative assessment model that pairs land use scenarios with representative con-332
centration pathways. For the current models, we used historical land-use maps (LUH2333
v2h), which are intended for use over the period 850 to 2015 C.E. 81 . To capture species’334
habitat preference, we collated data for all 3,870 mammal species from the IUCN Habitat335
Classification Scheme version 3.1. We then mapped 104 unique IUCN habitat classifi-336
cations onto the eight land use types present in the LUH dataset. For 962 species, no337
habitat data was available, or no correspondence existed between a land type in the IUCN338
scheme and our land use data; for these species, land use filters were not used. Filtering339
based on habitat was done conservatively: species were allowed in current and future340
ranges to exist in a pixel if any non-zero percent was assigned a suitable habitat type;341
almost all pixels contain multiple habitats. In some scenarios, human settlements cover342
at least some of a pixel for most of the world, allowing synanthropic species to persist343
throughout most of their climatically-suitable range. For those with habitat data, the344
average reduction in range from habitat filtering was 7.6% of pixels.345
11
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Refining the dataset346
Of the 3,870 species for which we generated distribution models, 103 were aquatic mam-347
mals (cetaceans, sirenians, pinnipeds, and sea otters), and 382 were not present in the348
mammalian supertree that we used for phylogenetic data82 . These species were ex-349
cluded. Aquatic species were removed using a two-filter approach, by cross-referencing350
with Pantheria83 . These results were verified by checking no species only had marine351
habitat use types (see ‘Habitat range and land use’). We also excluded 246 monotremes352
and marsupials because the shape of the supertree prevented us from fitting satisfactory353
GAMM smooths to the phylogeny effect, leaving 3,139 non-marine Eutherian mammals354
with associated phylogenetic data.355
Predicting future species distributions356
We modeled a total of 16 possible futures, produced by four paired climate-land use357
change pathways and two optional filters on species ranges (habitat preferences and dis-358
persal limits). The full matrix of possible scenarios captures a combination of scenario359
uncertainty about global change and epistemological uncertainty about how best to pre-360
dict species’ range shifts. By filtering possible future distributions based on climate, land361
use, and dispersal constraints, we aimed to maximize realism; our predictions were con-362
gruent with extensive prior literature on climate- and land use-driven range loss84,85,86.363
Climate and land use futures364
Species distribution models were projected for 2070 using climate models, and then spa-365
tially filtered by land use projections. Climate and land-use future pathways are coupled366
by the Land Use Harmonization 2.0 integrative assessment model87,81, such that every367
future has a representative concentration pathway (RCP) for climate and a shared so-368
cioeconomic pathway (SSP) for land use. For climate we used the HadGEM2 Earth369
System Model projections for 2070, with the four standard RCPs: 2.6, 4.5, 6.0, and 8.5370
(where the values represent added W/m2of solar radiation by the end of the century371
due to greenhouse gas emissions). These were respectively paired with SSP 1 (“Sustain-372
ability”); SSP 2 (“Middle of the Road”); SSP 4 (“Inequality”); and SSP 5 (“Fossil-Fueled373
Development”).374
These pairings can be thought of as a gradient of scenarios of global change with differ-375
ent levels of severity and sustainability. Not all scenarios are possible; the four we selected376
are drawn as some of the most representative from an underlying “scenario matrix” that377
includes every possible parameterization, some of which are entirely incompatible88 . (For378
example, in the vast majority of integrative assessment models, decarbonization cannot379
12
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
be achieved fast enough in the SSP 5 scenario to achieve RCP 2.6.) As pairs, SSP-RCP380
narratives can be merged to create overall narratives about how global change could look.381
For example, in SSP 1-RCP 2.6, a global transition to renewable energy and mitigation of382
climate change corresponds to sustainable population growth and economic development.383
Driven by international cooperation on climate agreements, afforestation and bioenergy384
cropland become major land uses, while tropical deforestation is strongly reduced. In385
contrast, in SSP 5-RCP 8.5, business-as-usual development leads to catastrophic levels386
of warming, unsustainable population growth and increasing poverty, and massive land387
conversion89,90 .388
Limiting dispersal capacity389
Not all species can disperse to all environments, and not all species have equal disper-390
sal capacity—in ways likely to covary with viral sharing properties. We follow a rule391
proposed by Schloss et al.31 , who described an approximate formula for mammal range392
shift capacity based on body mass and trophic position. For carnivores, the maximum393
distance traveled in a generation is given as D= 40.7M0.81, where Dis distance in kilo-394
meters and Mis body mass in kilograms. For herbivores and omnivores, the maximum395
is estimated as D= 3.31M0.65.396
We used mammalian diet data from the EltonTraits database91, and used the same397
cutoff as Schloss to identify carnivores as any species with 10% or less plants in their398
diet. We used body mass data from EltonTraits in the Schloss formula to estimate399
maximum generational dispersal, and converted estimates to annual maximum dispersal400
rates by dividing by generation length, as previously estimated by another comprehensive401
mammal dataset92. We multiply by 50 years and use the resulting distance as a buffer402
around the original range map, and constrain possible range shifts within that buffer. For403
420 species with missing data in one of the required sources, we interpolated dispersal404
distance based on the closest relative in our supertree with a dispersal velocity estimate.405
Qualified by the downsides of assuming full dispersal93 , we excluded bats from the406
assumed scaling of dispersal limitations. The original study by Schloss et al.31 chose407
to omit bats entirely, and subsequent work has not proposed any alternative formula.408
Moreover, the Schloss formula performs notably poorly for bats: for example, it would409
assign the largest bat in our study, the Indian flying fox (Pteropus giganteus), a disper-410
sal capacity lower than that of the gray dwarf hamster (Cricetulus migratorius ). Bats411
were instead given full dispersal in all scenarios: given significant evidence that some bat412
species regularly cover continental distances 43,44 , and that isolation by distance is uncom-413
mon within many bats’ ranges46 , we felt this was a defensible assumption for modeling414
purposes. Moving forward, the rapid range shifts already observed in many bat species415
13
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
(see main text) could provide an empirical reference point to fit a new allometric scaling416
curve (after standardizing those results for the studies’ many different methodologies).417
A different set of functional traits likely govern the scaling of bat dispersal, chiefly the418
aspect ratio (length:width) of wings, which is a strong predictor of population genetic419
differentiation46 . Migratory status would also be important to include as a predictor420
although here, we exclude information on long-distance migration for all species (due to421
a lack of any real framework for adding that information to species distribution models422
in the literature).423
Explaining spatial patterns424
To explore the geography of novel assemblages, we used linear models which predicted the425
number of first encounters (novel overlap of species pairs) at the 25km level (N= 258,539426
grid cells). Explanatory variables included: richness (number of species inhabiting the427
grid cell in our predicted current ranges for the given scenario); elevation in meters (de-428
rived from the US Geological Service Global Multi-resolution Terrain Elevation Data429
2010 dataset); and the predominant land cover type for the grid cell. We simplified430
the classification scheme for land use types into five categories for these models (human431
settlement, cropland, rangeland and pasture, forest, and unforested wildland), and as-432
signed pixels a single land use type based on the maximum probability from the land433
use scenarios. We fitted a model for each scenario and pair of biological assumptions;434
because of the large effect bats had on the overall pattern, we retrained these models on435
subsets of encounters with and without a bat species involved. To help model fitting, we436
log(x+1)-transformed the response variable (number of overlaps in the pixel) and both437
continuous explanatory variables (meters of elevation above the lowest point and species438
richness). Because some elevation values were lower than 0 (i.e., below sea level), we439
treated elevation as meters above the lowest terrestrial point rather than meters above440
sea level to allow us to log-transform the data.441
Viral sharing models442
Generalized Additive Mixed Models443
We used a previously-published model of the phylogeography of viral sharing patterns444
to make predictions of future viral sharing17 . This model was based on an analysis of445
510 viruses shared between 682 mammal species3, and predicted the probability that a446
pair of mammal species will share a virus given their geographic range overlap and phy-447
logenetic relatedness. The original study uncovered strong, nonlinear effects of spatial448
overlap and phylogenetic similarity in determining viral sharing probability, and simu-449
14
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
lating the unobserved global network using these effect estimates capitulated multiple450
macroecological patterns of viral sharing.451
In the original study, a Generalized Additive Mixed Model (GAMM) was used to452
predict virus sharing as a binary variable, based on (1) geographic range overlap; (2) phy-453
logenetic similarity; and (3) species identity as a multi-membership random effect. The454
phylogeographic explanatory variables were obtained from two broadly available, low-455
resolution data sources: pairwise phylogenetic similarity was derived from a mammalian456
supertree previously modified for host-pathogen studies 82,3 , with similarity defined as457
the inverse of the cumulative branch length between two species, scaled to between 0458
and 1. Geographic overlap was defined as the area of overlap between two species’ IUCN459
range maps, divided by their cumulative range size94.460
We first retrained the GAMMs from 17 on the pairwise overlap matrix of species distri-461
bution models generated for this study, so that present predictions would be comparable462
with future distributions. Of the 3,139 species in our reduced dataset, 544 had viral463
records in our viral sharing dataset and shared with at least one other mammal, and464
were used to retrain the GAMM from17 . To check the performance of the GAMM, we465
predicted sharing patterns with a) only random effects, b) only fixed effects, and c) with466
both. Although species-level random effects had a mean effect of 0, excluding them en-467
tirely resulted in a substantial underestimation of the mean viral sharing rates across the468
network (mean sharing 0.02 compared to 0.06). Therefore to ensure that the model469
recapitulated traits of the observed network, we simulated 1,000 binary sharing networks470
when predicting with only fixed effects, randomly drawing species-level random effects471
in each iteration. The mean sharing value across these iterations closely approximated472
observed sharing probability (0.06).473
Model validation and limits474
Compared to the current viral sharing matrix, the model performs well with only fixed475
effects (AUC = 0.80) and extremely well with both fixed and random effects (AUC =476
0.93). The model explained a very similar proportion of the deviance in viral sharing to477
that in Albery et al. 17 (44.5% and 44.8% respectively).478
In practice, several unpredictable but confounding factors could affect the reliability479
of this model as a forecasting tool, including temperature sensitivity of viral evolution in480
host jumps68, or increased susceptibility of animals with poorer health in lower-quality481
habitat or unfavorable climates. Moreover, once viruses can produce an infection, their482
ability to transmit within a new species is an evolutionary race between mutation and483
recombination rates in viral genomes, host innate and adaptive immunity, virulence-484
related mortality, and legacy constraints of coevolution with prior hosts and vectors60,61.485
15
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
But data cataloging these precise factors are hardly comprehensive for the hundreds of486
zoonotic viruses, let alone for the thousands of undescribed viruses in wildlife. Moreover,487
horizontal transmission is not necessary for spillover potential to be considered significant;488
for example, viruses like rabies or West Nile virus are not transmitted within human489
populations but humans are still noteworthy hosts.490
Mapping opportunities for sharing491
We used the GAMM effect estimates to predict viral sharing patterns across the 3,139492
mammals with associated geographic range and phylogenetic data, for both the present493
and future scenarios. By comparing current and future sharing probabilities for each of494
the four global change scenarios, we estimated which geographic and taxonomic patterns495
of viral sharing would likely emerge. We separately examined patterns of richness, pat-496
terns of sharing probability, and their change (i.e., future sharing probability - current497
sharing probability, giving the expected probability of a novel sharing event).498
A subset of the mammals in our dataset were predicted to encounter each other for the499
first time during range shifts. For each of these pairwise first encounters, we extracted the500
area of overlap in every future scenario, and assigned each overlap a probability of sharing501
from the mean GAMM predictions and mapped the mean and cumulative probability of502
a new sharing event happening in a given geographic pixel.503
Case study on Zaire ebolavirus504
For a case study in possible significant cross-species transmission, we compiled a list505
of known hosts of Zaire ebolavirus (ZEBOV), a zoonosis with high host breadth that506
has been known to cause wildlife die-offs, but has no known definitive reservoir. Hosts507
were taken from two sources: the training dataset on host-virus associations 3, and an508
additional dataset of filovirus testing in bats30 . In the latter case, any bats that have509
been reported antibody positive or PCR-positive for ZEBOV were included. A total510
of 13 current “known hosts” in Africa were used to predict current possible hosts, and511
first encounters in all scenarios. We restricted our analysis to Africa because there is512
no published evidence that Zaire ebolavirus actively circulates outside Africa; although513
some bat species outside Africa have tested positive for antibodies to ZEBOV, this is514
likely due to cross-reactivity with other undiscovered filoviruses95,96,30.515
Overlap with human populations516
To examine the possibility that hotspots of cross-species transmission would overlap with517
human populations, we used SEDAC’s global population projections version 1.0 for the518
16
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
year 207097 . We aggregated these to native resolution, for each of the four SSP paired519
with the native RCP/SSP pairing for the species distribution models. In Figure 4 we520
present the population projections for SSP 1, which pairs with RCP 2.6.521
17
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Acknowledgements522
This paper is the culmination of several years of idea development and owes special thanks523
to many people, including the entire Bansal Lab, Laura Ward Alexander, Kevin Burgio,524
Eric Dougherty, Romain Garnier, Wayne Getz, Peta Hitchens, Christine Johnson, and525
Isabel Ott. We especially thank Laura Alexander for sharing bat filovirus testing sources526
used to compile the Ebola sub-network. Thanks are also extended to José Hidasi-Neto for527
publicly-available data visualization code. This research was supported by the George-528
town Environment Initiative and the National Socio-Environmental Synthesis Center529
(SESYNC) under funding received from the National Science Foundation DBI-1639145.530
C.M. acknowledges funding from National Science Foundation grant DBI-1913673.531
Author Contributions532
CJC and GFA conceived the study. CM, CJC, and CHT developed species distribution533
models; GFA, EAE, KJO, and NR developed the generalized additive models. CJC, GFA,534
and CMZ integrated the predictions of species distributions and viral sharing patterns535
and designed visualizations. All authors contributed to the writing of the manuscript.536
18
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Figures537
A.
B.
Figure 1: Climate change will drive novel viral sharing among mammal species.
The projected number of novel viral sharing events among mammal species in 2070 based
on host species geographic range shifts from climate change (RCP 2.6) and land-use
change (SSP 1), without dispersal limits (A) and with dispersal limitation (B).
19
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
A.
B.
C.
D.
E.
Figure 2: Bats disproportionately drive future novel viral sharing. The spatial
pattern of first encounters differs among range-shifting mammal pairs including bat-
bat and bat-nonbat encounters (A) and only encounters among non-bats (B). Using a
linear model, we show that elevation (C), species richness (D), and land use (E) together
explain 57.7% of deviance in new overlaps for bats, and 25.8% for non-bats. Slopes
for the elevation effect were generally steeply positive: a log10 -increase in elevation was
associated with between a 0.4-1.41 log10-increase in first encounters.
20
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
C.
D.
A. B.
Figure 3: Range expansions will expose naive hosts to zoonotic reservoirs. (A)
The predicted distribution of known African hosts of Zaire ebolavirus. (B) The change in
richness of these hosts as a result of range shifts. (C) Projected first encounters with non-
Ebola hosts. (D) Bat-primate first encounters are projected to occur globally, producing
novel sharing events.
21
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Figure 4: Novel viral sharing events coincide with population centers. In 2070
(RCP 2.6; climate only), human population centers in equatorial Africa, south China and
southeast Asia will overlap with projected hotspots of cross-species viral transmission in
wildlife. (Both variables are linearly rescaled to 0 to 1.)
22
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Degree
Extended Data Figure 1: The mammal-virus network. The present-day viral sharing
network by mammal order inferred from modeled pairwise predictions of viral sharing
probabilities. Edge width denotes the expected number of shared viruses (the sum of
pairwise species-species viral sharing probabilities), with most sharing existing among
the most speciose and closely-related groups. Edges shown in the network are the top
25% of links. Nodes are sized by total number of species in that order in the host-virus
association dataset, color is scaled by degree.
23
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 2: Predicted phylogeographic structure of viral sharing.
Phylogeographic prediction of viral sharing using a generalized additive mixed model. Vi-
ral sharing increases as a function of phylogenetic similarity (A) and geographic overlap
(B), fit together as a tensor interaction (C). White contour lines denote 10% increments of
sharing probability. Declines at high values of overlap may be an artefact of model struc-
ture and low sampling in the upper levels of geographic overlap, shown in a hexagonal
bin chart for raw data (D).
24
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 3: Outcomes by model formulation and climate change
scenario. Heatmaps displaying predicted changes across model formulations. (A) Range
expansions were highest in non-dispersal-limited scenarios and in milder RCPs. (B) The
number of predicted first encounters was higher in non-dispersal-limited scenarios and in
milder RCPs. (C) The number of expected new viral sharing events was higher in non-
dispersal-limited scenarios and in more severe RCPs. (D) The overall change in sharing
probability (connectance) across the viral sharing network between the present day and
the future scenarios; absolute change is minimal but positive across all scenarios, being
greatest in non-dispersal-limited scenarios and in milder RCPs.
25
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 4: Geographic distribution of first encounters. Predictions
were carried out for four representative concentration pathways (RCPs), accounting for
climate change and land use change, without (left) and with dispersal limits (right).
Darker colours correspond to greater numbers of first encounters in the pixel.
26
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 5: Geographic distribution of expected viral sharing
events from first encounters. Predictions were carried out for future distributions for
four representative concentration pathways (RCPs), accounting for climate change and
land use change, without (left) and with dispersal limits (right). Darker colours corre-
spond to greater numbers of new viral sharing events in the pixel. Probability of new
viral sharing was calculated by subtracting the species pair’s present sharing probability
from their sharing probability that our viral sharing GAMMs predicted. This probability
was projected across the species pair’s range intersection, and then summed across all
novel species pairs in each pixel.
27
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 6: Order-level heterogeneity in first encounters. Dispersal
stratifies the number of first encounters (RCP 2.6 with all range filters), where some
orders have more than expected at random, based on the mean number of first encounters
and order size (line).
28
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 7: Projected viral sharing from suspected Ebola reservoirs
is dominated by bats. Node size is proportional to (left) the number of suspected Ebola
host species in each order, which connect to (middle) first encounters with potentially
naive host species; and (right) the number of projected viral sharing events in each
receiving group. (Node size denotes proportions out of 100% within each column total.)
While Ebola hosts will encounter a much wider taxonomic range of mammal groups than
current reservoirs, the vast majority of viral sharing will occur disproportionately in bats.
29
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 8: Data processing workflow. Summary of species inclusion
across the modeling pipeline for species distributions and viral sharing models. The final
analyses in the main text use 3,139 species of Eutherian mammals across all scenarios.
30
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
Extended Data Figure 9: Species distribution modeling workflow for a single
species. A focal species (the European red deer, Cervus elaphus ) is displayed as an
illustrative example. The present day climate prediction (top left) was clipped to the
same continent according to the IUCN distribution (top right). This was then clipped
according to Cervus elaphus land use (second row, left). The known dispersal distance of
the red deer was used to buffer the climate distribution (second row, right). The future
distribution predictions (RCP 2.6 shown as an example) are displayed in the bottom
four panels, for each of the four pipelines: only climate (third row, left); climate +
dispersal clip (third row, right); climate + land use clip (bottom row, left) and climate
+ land use + dispersal clip (bottom row, right). The four distributions clearly display
the limiting effect of the dispersal filter (bottom right panels) in reducing the probability
of novel species interactions (bottom left panels). The land use clip had little effect on
this species as the entire distribution area was habitable for the red deer.
31
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
References538
1. Carlson, C. J, Zipfel, C. M, Garnier, R, & Bansal, S. (2019) Global estimates539
of mammalian viral biodiversity accounting for host sharing. Nature Ecology and540
Evolution 3, 1070–1075.541
2. Carroll, D, Daszak, P, Wolfe, N. D, Gao, G. F, Morel, C. M, Morzaria, S, Pablos-542
Méndez, A, Tomori, O, & Mazet, J. A. (2018) The global virome project. Science543
359, 872–874.544
3. Olival, K. J, Hosseini, P. R, Zambrana-Torrelio, C, Ross, N, Bogich, T. L, & Daszak,545
P. (2017) Host and viral traits predict zoonotic spillover from mammals. Nature546
546, 646.547
4. Altizer, S, Ostfeld, R. S, Johnson, P. T, Kutz, S, & Harvell, C. D. (2013) Climate548
change and infectious diseases: from evidence to a predictive framework. Science549
341, 514–519.550
5. Cleaveland, S, Haydon, D, & Taylor, L. (2007) Overviews of pathogen emergence:551
which pathogens emerge, when and why? Wildlife and Emerging Zoonotic Diseases:552
The Biology, Circumstances and Consequences of Cross-Species Transmission pp.553
85–111.554
6. Parrish, C. R, Holmes, E. C, Morens, D. M, Park, E.-C, Burke, D. S, Calisher, C. H,555
Laughlin, C. A, Saif, L. J, & Daszak, P. (2008) Cross-species virus transmission and556
the emergence of new epidemic diseases. Microbiology and Molecular Biology Reviews557
72, 457–470.558
7. Woolhouse, M. E, Haydon, D. T, & Antia, R. (2005) Emerging pathogens: the559
epidemiology and evolution of species jumps. Trends in Ecology & Evolution 20,560
238–244.561
8. Hoberg, E. P & Brooks, D. R. (2015) Evolution in action: climate change, biodi-562
versity dynamics and emerging infectious disease. Philosophical Transactions of the563
Royal Society B: Biological Sciences 370, 20130553.564
9. Burrows, M. T, Schoeman, D. S, Richardson, A. J, Molinos, J. G, Hoffmann, A,565
Buckley, L. B, Moore, P. J, Brown, C. J, Bruno, J. F, Duarte, C. M, et al. (2014)566
Geographical limits to species-range shifts are suggested by climate velocity. Nature567
507, 492.568
32
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
10. Chen, I.-C, Hill, J. K, Ohlemüller, R, Roy, D. B, & Thomas, C. D. (2011) Rapid569
range shifts of species associated with high levels of climate warming. Science 333,570
1024–1026.571
11. Carlson, C. J, Burgio, K. R, Dougherty, E. R, Phillips, A. J, Bueno, V. M, Clements,572
C. F, Castaldo, G, Dallas, T. A, Cizauskas, C. A, Cumming, G. S, et al. (2017) Par-573
asite biodiversity faces extinction and redistribution in a changing climate. Science574
Advances 3, e1602422.575
12. Wolfe, N. D, Dunavan, C. P, & Diamond, J. (2007) Origins of major human infectious576
diseases. Nature 447, 279.577
13. Lloyd-Smith, J. O, George, D, Pepin, K. M, Pitzer, V. E, Pulliam, J. R, Dobson,578
A. P, Hudson, P. J, & Grenfell, B. T. (2009) Epidemic dynamics at the human-animal579
interface. Science 326, 1362–1367.580
14. Plowright, R. K, Parrish, C. R, McCallum, H, Hudson, P. J, Ko, A. I, Graham,581
A. L, & Lloyd-Smith, J. O. (2017) Pathways to zoonotic spillover. Nature Reviews582
Microbiology 15, 502.583
15. Gould, S. J & Lewontin, R. C. (1979) The spandrels of San Marco and the Panglos-584
sian paradigm: a critique of the adaptationist programme. Proceedings of the Royal585
Society of London. Series B: Biological Sciences 205, 581–598.586
16. Alexander, K. A, Carlson, C. J, Lewis, B. L, Getz, W. M, Marathe, M. V, Eubank,587
S. G, Sanderson, C. E, & Blackburn, J. K. (2018) The ecology of pathogen spillover588
and disease emergence at the human-wildlife-environment interface. The Connections589
Between Ecology and Infectious Disease pp. 267–298.590
17. Albery, G. F, Eskew, E. A, Ross, N, & Olival, K. J. (2019) Predicting the global591
mammalian viral sharing network using phylogeography. bioRxiv p. 732255.592
18. Faria, N. R, Suchard, M. A, Rambaut, A, Streicker, D. G, & Lemey, P. (2013) Simul-593
taneously reconstructing viral cross-species transmission history and identifying the594
underlying constraints. Philosophical Transactions of the Royal Society B: Biological595
Sciences 368, 20120196.596
19. Dougherty, E. R, Seidel, D. P, Carlson, C. J, Spiegel, O, & Getz, W. M. (2018)597
Going through the motions: incorporating movement analyses into disease research.598
Ecology Letters 21, 588–604.599
20. Sun, B, Jia, L, Liang, B, Chen, Q, & Liu, D. (2018) Phylogeography, transmission,600
and viral proteins of Nipah virus. Virologica Sinica 33, 385–393.601
33
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
21. Huang, S, Bininda-Emonds, O. R, Stephens, P. R, Gittleman, J. L, & Altizer, S.602
(2014) Phylogenetically related and ecologically similar carnivores harbour similar603
parasite assemblages. Journal of Animal Ecology 83, 671–680.604
22. Davies, T. J & Pedersen, A. B. (2008) Phylogeny and geography predict pathogen605
community similarity in wild primates and humans. Proceedings of the Royal Society606
B: Biological Sciences 275, 1695–1701.607
23. Farrell, M. J & Davies, T. J. (2019) Disease mortality in domesticated animals is608
predicted by host evolutionary relationships. Proceedings of the National Academy609
of Sciences 116, 7911–7915.610
24. Longdon, B, Brockhurst, M. A, Russell, C. A, Welch, J. J, & Jiggins, F. M. (2014)611
The evolution and genetics of virus host shifts. PLoS Pathogens 10, e1004395.612
25. Streicker, D. G, Turmelle, A. S, Vonhof, M. J, Kuzmin, I. V, McCracken, G. F,613
& Rupprecht, C. E. (2010) Host phylogeny constrains cross-species emergence and614
establishment of rabies virus in bats. Science 329, 676–679.615
26. Jones, K. E, Patel, N. G, Levy, M. A, Storeygard, A, Balk, D, Gittleman, J. L, &616
Daszak, P. (2008) Global trends in emerging infectious diseases. Nature 451, 990.617
27. Allen, T, Murray, K. A, Zambrana-Torrelio, C, Morse, S. S, Rondinini, C, Di Marco,618
M, Breit, N, Olival, K. J, & Daszak, P. (2017) Global hotspots and correlates of619
emerging zoonotic diseases. Nature Communications 8, 1124.620
28. Johnson, C. K, Hitchens, P. L, Evans, T. S, Goldstein, T, Thomas, K, Clements,621
A, Joly, D. O, Wolfe, N. D, Daszak, P, Karesh, W. B, et al. (2015) Spillover and622
pandemic properties of zoonotic viruses with high host plasticity. Scientific Reports623
5, 14830.624
29. Han, B. A, Kramer, A. M, & Drake, J. M. (2016) Global patterns of zoonotic disease625
in mammals. Trends in Parasitology 32, 565–577.626
30. Han, B. A, Schmidt, J. P, Alexander, L. W, Bowden, S. E, Hayman, D. T, & Drake,627
J. M. (2016) Undiscovered bat hosts of filoviruses. PLoS Neglected Tropical Diseases628
10, e0004815.629
31. Schloss, C. A, Nuñez, T. A, & Lawler, J. J. (2012) Dispersal will limit ability of630
mammals to track climate change in the western hemisphere. Proceedings of the631
National Academy of Sciences 109, 8606–8611.632
34
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
32. Loarie, S. R, Duffy, P. B, Hamilton, H, Asner, G. P, & Field, C. B. (2009) The633
velocity of climate change. Nature 462, 1052–1055.634
33. Hickling, R, Roy, D. B, Hill, J. K, Fox, R, & Thomas, C. D. (2006) The distributions635
of a wide range of taxonomic groups are expanding polewards. Global Change Biology636
12, 450–455.637
34. Mason, S. C, Palmer, G, Fox, R, Gillings, S, Hill, J. K, Thomas, C. D, & Oliver,638
T. H. (2015) Geographical range margins of many taxonomic groups continue to639
shift polewards. Biological Journal of the Linnean Society 115, 586–597.640
35. Bebber, D. P, Ramotowski, M. A, & Gurr, S. J. (2013) Crop pests and pathogens641
move polewards in a warming world. Nature Climate Change 3, 985.642
36. Davidson, R, Simard, M, Kutz, S. J, Kapel, C. M, Hamnes, I. S, & Robertson, L. J.643
(2011) Arctic parasitology: why should we care? Trends in Parasitology 27, 239–245.644
37. Hoberg, E. P, Cook, J, Agosta, S, Boeger, W, Galbreath, K, Laaksonen, S, Kutz, S,645
& Brooks, D. (2017) Arctic systems in the quaternary: ecological collision, faunal646
mosaics and the consequences of a wobbling climate. Journal of Helminthology 91,647
409–421.648
38. Trisos, C. H, Merow, C, & Pigot, A. L. (2020) Timing and abruptness of ecological649
disruption from climate change. Nature p. in press.650
39. Colwell, R. K, Brehm, G, Cardelús, C. L, Gilman, A. C, & Longino, J. T. (2008)651
Global warming, elevational range shifts, and lowland biotic attrition in the wet652
tropics. Science 322, 258–261.653
40. Pauchard, A, Milbau, A, Albihn, A, Alexander, J, Burgess, T, Daehler, C, Englund,654
G, Essl, F, Evengård, B, Greenwood, G. B, et al. (2016) Non-native and native655
organisms moving into high elevation and high latitude ecosystems in an era of656
climate change: new challenges for ecology and conservation. Biological Invasions657
18, 345–353.658
41. Atkinson, C. T & LaPointe, D. A. (2009) Introduced avian diseases, climate change,659
and the future of hawaiian honeycreepers. Journal of Avian Medicine and Surgery660
23, 53–63.661
42. Schloss, A. L, Kicklighter, D. W, Kaduk, J, Wittenberg, U, & Intercomparison,662
T. P. O. T. P. N. M. (1999) Comparing global models of terrestrial net primary663
productivity (NPP): comparison of NPP to climate and the Normalized Difference664
Vegetation Index (NDVI). Global Change Biology 5, 25–34.665
35
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
43. Peel, A. J, Sargan, D. R, Baker, K. S, Hayman, D. T, Barr, J. A, Crameri, G, Suu-666
Ire, R, Broder, C. C, Lembo, T, Wang, L.-F, et al. (2013) Continent-wide panmixia667
of an African fruit bat facilitates transmission of potentially zoonotic viruses. Nature668
Communications 4, 2770.669
44. Riesle-Sbarbaro, S. A, Amponsah-Mensah, K, De Vries, S, Nicolas, V, Lalis, A, Suu-670
Ire, R, Cunningham, A. A, Wood, J. L, & Sargan, D. R. (2018) The Gambian671
epauletted fruit bat shows increased genetic divergence in the Ethiopian highlands672
and in an area of rapid urbanization. Ecology and Evolution 8, 12803–12820.673
45. Wu, J. (2016) Detection and attribution of the effects of climate change on bat674
distributions over the last 50 years. Climatic Change 134, 681–696.675
46. Olival, K. (2012) Evolutionary and ecological correlates of population genetic struc-676
ture in bats. Evolutionary history of bats fossils, molecules and morphology pp.677
267–316.678
47. Olival, K. J, Latinne, A, Islam, A, Epstein, J. H, Hersch, R, Engstrand, R. C, Gurley,679
E. S, Amato, G, Luby, S. P, & Daszak, P. (2019) Population genetics of fruit bat680
reservoir informs the dynamics, distribution and diversity of Nipah virus. Molecular681
Ecology p. in press.682
48. Ancillotto, L, Santini, L, Ranc, N, Maiorano, L, & Russo, D. (2016) Extraordi-683
nary range expansion in a common bat: the potential roles of climate change and684
urbanisation. The Science of Nature 103, 15.685
49. Ancillotto, L, Budinski, I, Nardone, V, Di Salvo, I, Della Corte, M, Bosso, L, Conti,686
P, & Russo, D. (2018) What is driving range expansion in a common bat? hints687
from thermoregulation and habitat selection. Behavioural Processes 157, 540–546.688
50. Geluso, K, Mollhagen, T. R, Tigner, J. M, & Bogan, M. A. (2005) Westward expan-689
sion of the eastern pipistrelle (Pipistrellus subflavus) in the United States, including690
new records from New Mexico, South Dakota, and Texas. Western North American691
Naturalist 65, 12.692
51. Kurta, A, Winhold, L, Whitaker, J. O, & Foster, R. (2007) Range expansion and693
changing abundance of the eastern pipistrelle (Chiroptera: Vespertilionidae) in the694
central Great Lakes region. The American Midland Naturalist 157, 404–412.695
52. Lundy, M, Montgomery, I, & Russ, J. (2010) Climate change-linked range expan-696
sion of Nathusius’ pipistrelle bat, Pipistrellus nathusii (Keyserling & Blasius, 1839).697
Journal of Biogeography 37, 2232–2242.698
36
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
53. McCracken, G. F, Bernard, R. F, Gamba-Rios, M, Wolfe, R, Krauel, J. J, Jones,699
D. N, Russell, A. L, & Brown, V. A. (2018) Rapid range expansion of the Brazilian700
free-tailed bat in the southeastern United States, 2008–2016. Journal of Mammalogy701
99, 312–320.702
54. Roberts, B. J, Catterall, C. P, Eby, P, & Kanowski, J. (2012) Latitudinal range shifts703
in Australian flying-foxes: A re-evaluation. Austral Ecology 37, 12–22.704
55. Uhrin, M, Hüttmeir, U, Kipson, M, Estók, P, Sachanowicz, K, Bücs, S, Karapandža,705
B, Paunović, M, Presetnik, P, Bashta, A.-T, et al. (2016) Status of Savi’s pipistrelle706
Hypsugo savii (Chiroptera) and range expansion in Central and south-eastern Eu-707
rope: a review. Mammal Review 46, 1–16.708
56. Brook, C. E & Dobson, A. P. (2015) Bats as ‘special’ reservoirs for emerging zoonotic709
pathogens. Trends in Microbiology 23, 172–180.710
57. Luis, A. D, Hayman, D. T, O’Shea, T. J, Cryan, P. M, Gilbert, A. T, Pulliam,711
J. R, Mills, J. N, Timonin, M. E, Willis, C. K, Cunningham, A. A, et al. (2013) A712
comparison of bats and rodents as reservoirs of zoonotic viruses: are bats special?713
Proceedings of the Royal Society B: Biological Sciences 280, 20122753.714
58. Olival, K. J, Weekley, C. C, & Daszak, P. (2015) Are bats really ‘special’ as viral715
reservoirs? what we know and need to know. Bats and Viruses pp. 281–294.716
59. Wang, L.-F, Walker, P. J, & Poon, L. L. (2011) Mass extinctions, biodiversity and717
mitochondrial function: are bats ‘special’ as reservoirs for emerging viruses? Current718
Opinion in Virology 1, 649–657.719
60. Geoghegan, J. L, Senior, A. M, Di Giallonardo, F, & Holmes, E. C. (2016) Virological720
factors that increase the transmissibility of emerging human viruses. Proceedings of721
the National Academy of Sciences 113, 4170–4175.722
61. Walker, J. W, Han, B. A, Ott, I. M, & Drake, J. M. (2018) Transmissibility of723
emerging viral zoonoses. PloS One 13, e0206926.724
62. Pigott, D. M, Millear, A. I, Earl, L, Morozoff, C, Han, B. A, Shearer, F. M, Weiss,725
D. J, Brady, O. J, Kraemer, M. U, Moyes, C. L, et al. (2016) Updates to the zoonotic726
niche map of Ebola virus disease in Africa. Elife 5, e16412.727
63. Bermejo, M, Rodríguez-Teijeiro, J. D, Illera, G, Barroso, A, Vilà, C, & Walsh, P. D.728
(2006) Ebola outbreak killed 5000 gorillas. Science 314, 1564–1564.729
37
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
64. Hayman, D. T. (2019) African primates: Likely victims, not reservoirs, of730
ebolaviruses. The Journal of Infectious Diseases.731
65. Goldstein, T, Anthony, S. J, Gbakima, A, Bird, B. H, Bangura, J, Tremeau-Bravard,732
A, Belaganahalli, M. N, Wells, H. L, Dhanota, J. K, Liang, E, et al. (2018) The733
discovery of Bombali virus adds further support for bats as hosts of ebolaviruses.734
Nature Microbiology 3, 1084.735
66. Williams, J. N. (2013) Humans and biodiversity: population and demographic trends736
in the hotspots. Population and Environment 34, 510–523.737
67. Cheng, Y, Tjaden, N. B, Jaeschke, A, Lühken, R, Ziegler, U, Thomas, S. M, &738
Beierkuhnlein, C. (2018) Evaluating the risk for Usutu virus circulation in Europe:739
comparison of environmental niche models and epidemiological models. International740
Journal of Health Geographics 17, 35.741
68. Roberts, K. E, Hadfield, J. D, Sharma, M. D, & Longdon, B. (2018) Changes in742
temperature alter the potential outcomes of virus host shifts. PLoS Pathogens 14,743
e1007185.744
69. Faust, C. L, Dobson, A. P, Gottdenker, N, Bloomfield, L. S, McCallum, H. I, Gille-745
spie, T. R, Diuk-Wasser, M, & Plowright, R. K. (2017) Null expectations for disease746
dynamics in shrinking habitat: dilution or amplification? Philosophical Transactions747
of the Royal Society B: Biological Sciences 372, 20160173.748
70. Cunningham, A. (2018) Infectious disease threats to amphibian conservation. (Glas-749
gow Natural History Society), Vol. 27.750
71. VanWormer, E, Mazet, J, Hall, A, Gill, V, Boveng, P, London, J, Gelatt, T, Fadely,751
B, Lander, M, Sterling, J, et al. (2019) Viral emergence in marine mammals in the752
North Pacific may be linked to Arctic sea ice reduction. Scientific Reports 9, 1–11.753
72. Grubbs, F. E et al. (1950) Sample criteria for testing outlying observations. The754
Annals of Mathematical Statistics 21, 27–58.755
73. Renner, I. W, Elith, J, Baddeley, A, Fithian, W, Hastie, T, Phillips, S. J, Popovic,756
G, & Warton, D. I. (2015) Point process models for presence-only analysis. Methods757
in Ecology and Evolution 6, 366–379.758
74. Friedman, J, Hastie, T, & Tibshirani, R. (2010) Regularization paths for generalized759
linear models via coordinate descent. Journal of Statistical Software 33, 1.760
38
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
75. Phillips, S. J, Anderson, R. P, & Schapire, R. E. (2006) Maximum entropy modeling761
of species geographic distributions. Ecological Modelling 190, 231–259.762
76. Fick, S. E & Hijmans, R. J. (2017) WorldClim 2: new 1-km spatial resolution climate763
surfaces for global land areas. International Journal of Climatology 37, 4302–4315.764
77. Hastie, T, Tibshirani, R, & Friedman, J. H. (2009) The elements of statistical learn-765
ing: data mining, inference, and prediction. (Springer-Verlag, New York), 2nd edi-766
tion.767
78. Drake, J. M. (2015) Range bagging: a new method for ecological niche modelling768
from presence-only data. Journal of The Royal Society Interface 12, 20150086–9.769
79. Drake, J. M & Richards, R. L. (2018) Estimating environmental suitability. Ecosphere770
9, e02373.771
80. Hurtt, G. C, Chini, L. P, Frolking, S, Betts, R, Feddema, J, Fischer, G, Fisk, J,772
Hibbard, K, Houghton, R, Janetos, A, et al. (2011) Harmonization of land-use773
scenarios for the period 1500–2100: 600 years of global gridded annual land-use774
transitions, wood harvest, and resulting secondary lands. Climatic Change 109, 117.775
81. Hurtt, G, Chini, L, Sahajpal, R, Frolking, S, Bodirsky, B, Calvin, K, Doelman, J,776
Fisk, J, Fujimori, S, Goldewijk, K, et al. (2018) LUH2: Harmonization of global777
land-use scenarios for the period 850-2100.778
82. Fritz, S. A, Bininda-Emonds, O. R, & Purvis, A. (2009) Geographical variation in779
predictors of mammalian extinction risk: big is bad, but only in the tropics. Ecology780
Letters 12, 538–549.781
83. Jones, K. E, Bielby, J, Cardillo, M, Fritz, S. A, O’Dell, J, Orme, C. D. L, Safi,782
K, Sechrest, W, Boakes, E. H, Carbone, C, et al. (2009) PanTHERIA: a species-783
level database of life history, ecology, and geography of extant and recently extinct784
mammals: Ecological Archives E090-184. Ecology 90, 2648–2648.785
84. Jetz, W, Wilcove, D. S, & Dobson, A. P. (2007) Projected impacts of climate and786
land-use change on the global diversity of birds. PLoS Biology 5, e157.787
85. Pecl, G. T, Araújo, M. B, Bell, J. D, Blanchard, J, Bonebrake, T. C, Chen, I.-C,788
Clark, T. D, Colwell, R. K, Danielsen, F, Evengård, B, et al. (2017) Biodiversity789
redistribution under climate change: Impacts on ecosystems and human well-being.790
Science 355, eaai9214.791
39
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
86. Powers, R. P & Jetz, W. (2019) Global habitat loss and extinction risk of terrestrial792
vertebrates under future land-use-change scenarios. Nature Climate Change 9, 323.793
87. Hurtt, G. C, Chini, L. P, Frolking, S, Betts, R. A, Feddema, J, Fischer, G, Fisk, J. P,794
Hibbard, K, Houghton, R. A, Janetos, A, Jones, C. D, Kindermann, G, Kinoshita,795
T, Klein Goldewijk, K, Riahi, K, Shevliakova, E, Smith, S, Stehfest, E, Thomson,796
A, Thornton, P, van Vuuren, D. P, & Wang, Y. P. (2011) Harmonization of land-797
use scenarios for the period 1500–2100: 600 years of global gridded annual land-use798
transitions, wood harvest, and resulting secondary lands. Climatic Change 109,799
117–161.800
88. van Vuuren, D. P, Riahi, K, Calvin, K, Dellink, R, Emmerling, J, Fujimori, S, Kc, S,801
Kriegler, E, & O’Neill, B. (2017) The Shared Socio-economic Pathways: Trajectories802
for human development and global environmental change. Global Environmental803
Change 42, 148–152.804
89. Popp, A, Calvin, K, Fujimori, S, Havlik, P, Humpenöder, F, Stehfest, E, Bodirsky,805
B. L, Dietrich, J. P, Doelmann, J. C, Gusti, M, et al. (2017) Land-use futures in the806
shared socio-economic pathways. Global Environmental Change 42, 331–345.807
90. Riahi, K, van Vuuren, D. P, Kriegler, E, Edmonds, J, O’Neill, B. C, Fujimori, S,808
Bauer, N, Calvin, K, Dellink, R, Fricko, O, Lutz, W, Popp, A, Cuaresma, J. C, Kc,809
S, Leimbach, M, Jiang, L, Kram, T, Rao, S, Emmerling, J, Ebi, K, Hasegawa, T,810
Havlik, P, Humpenöder, F, Da Silva, L. A, Smith, S, Stehfest, E, Bosetti, V, Eom,811
J, Gernaat, D, Masui, T, Rogelj, J, Strefler, J, Drouet, L, Krey, V, Luderer, G,812
Harmsen, M, Takahashi, K, Baumstark, L, Doelman, J. C, Kainuma, M, Klimont, Z,813
Marangoni, G, Lotze-Campen, H, Obersteiner, M, Tabeau, A, & Tavoni, M. (2017)814
The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas815
emissions implications: An overview. Global Environmental Change 42, 153–168.816
91. Wilman, H, Belmaker, J, Simpson, J, de la Rosa, C, Rivadeneira, M. M, & Jetz,817
W. (2014) Eltontraits 1.0: Species-level foraging attributes of the world’s birds and818
mammals: Ecological archives e095-178. Ecology 95, 2027–2027.819
92. Pacifici, M, Santini, L, Di Marco, M, Baisero, D, Francucci, L, Marasini, G. G,820
Visconti, P, & Rondinini, C. (2013) Generation length for mammals. Nature Con-821
servation 5, 89.822
93. Bateman, B. L, Murphy, H. T, Reside, A. E, Mokany, K, & VanDerWal, J. (2013)823
Appropriateness of full-, partial-and no-dispersal scenarios in climate change impact824
modelling. Diversity and Distributions 19, 1224–1234.825
40
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
94. Araújo, M. B, Rozenfeld, A, Rahbek, C, & Marquet, P. A. (2011) Using species co-826
occurrence networks to assess the impacts of climate change. Ecography 34, 897–908.827
95. Olival, K. J, Islam, A, Yu, M, Anthony, S. J, Epstein, J. H, Khan, S. A, Khan, S. U,828
Crameri, G, Wang, L.-F, Lipkin, W. I, et al. (2013) Ebola virus antibodies in fruit829
bats, Bangladesh. Emerging Infectious Diseases 19, 270.830
96. Yang, X.-L, Zhang, Y.-Z, Jiang, R.-D, Guo, H, Zhang, W, Li, B, Wang, N, Wang,831
L, Waruhiu, C, Zhou, J.-H, et al. (2017) Genetically diverse filoviruses in Rousettus832
and Eonycteris spp. bats, China, 2009 and 2015. Emerging Infectious Diseases 23,833
482.834
97. Gao, J. (2017) Downscaling global spatial population projections from 1/8-degree835
to 1-km grid cells, (NCAR technical note NCAR/TN-537+ STR. Boulder, Colorado:836
National Center for Atmospheric Research), Technical report.837
41
All rights reserved. No reuse allowed without permission.
perpetuity.
this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for. http://dx.doi.org/10.1101/2020.01.24.918755doi: bioRxiv preprint first posted online Jan. 25, 2020;
... Climate change is often predicted as enhancing the transmission of mosquitoborne disease, although the consequences are often nonlinear (Reiter, 2001;Mordecai et al., 2013;Franklinos et al., 2019). The global ecology of mosquito and host can be used to better predict future geographic distributions of disease risk (Mordecai et al., 2013;Mordecai et al., 2019;Shapiro et al., 2017;Depinay et al., 2004;Ryan et al., 2019;Carlson et al., 2021). ...
Article
Full-text available
Parasites regularly switch into new host species, representing a disease burden and conservation risk to the hosts. The distribution of these parasites also gives insight into characteristics of ecological networks and genetic mechanisms of host-parasite interactions. Some parasites are shared across many species, whereas others tend to be restricted to hosts from a single species. Understanding the mechanisms producing this distribution of host specificity can enable more effective interventions and potentially identify genetic targets for vaccines or therapies. As ecological connections between human and local animal populations increase, the risk to human and wildlife health from novel parasites also increases. Which of these parasites will fizzle out and which have the potential to become widespread in humans? We consider the case of primate malarias, caused by Plasmodium parasites, to investigate the interacting ecological and evolutionary mechanisms that put human and nonhuman primates at risk for infection. Plasmodium host switching from nonhuman primates to humans led to ancient introductions of the most common malaria-causing agents in humans today, and new parasite switching is a growing threat, especially in Asia and South America. Based on a wild host-Plasmodium occurrence database, we highlight geographic areas of concern and potential areas to target further sampling. We also discuss methodological developments that will facilitate clinical and field-based interventions to improve human and wildlife health based on this eco-evolutionary perspective.
... A s the rate of infectious disease emergence continues to rise, it is becoming increasingly important to identify and understand the drivers of zoonotic risk in wild animals [1][2][3] . Humans are rapidly altering patterns of wildlife disease through a combination of climate change and land conversion, both of which are expected to drive increased spillover (that is, interspecific transmission of parasites from animals into humans [2][3][4][5][6][7][8]. Urban environments in particular are expected to facilitate the emergence of zoonotic pathogens in wildlife 3,7,[9][10][11] because of a combination of impaired immune systems fed by anthropogenic resources 10,12 and greater pollution 13 as well as increased proximity of wild animals to humans 7,14 . ...
Article
Full-text available
The world is rapidly urbanizing, inviting mounting concern that urban environments will experience increased zoonotic disease risk. Urban animals could have more frequent contact with humans, therefore transmitting more zoonotic parasites; however, this relationship is complicated by sampling bias and phenotypic confounders. Here we test whether urban mammal species host more zoonotic parasites, investigating the underlying drivers alongside a suite of phenotypic, taxonomic and geographic predictors. We found that urban-adapted mammals have more documented parasites and more zoonotic parasites: despite comprising only 6% of investigated species, urban mammals provided 39% of known host–parasite combinations. However, contrary to predictions, much of the observed effect was attributable to parasite discovery and research effort rather than to urban adaptation status, and urban-adapted species in fact hosted fewer zoonotic parasites than expected on the basis of their total parasite richness. We conclude that extended historical contact with humans has had a limited impact on zoonotic parasite richness in urban-adapted mammals; instead, their greater observed zoonotic richness probably reflects sampling bias arising from proximity to humans, supporting a near-universal conflation between zoonotic risk, research effort and synanthropy. These findings underscore the need to resolve the mechanisms linking anthropogenic change, sampling bias and observed wildlife disease dynamics. After compiling literature data on mammal parasites across urban and non-urban areas, the authors show that mammals in urban areas have more parasites overall without disproportionately more zoonotic ones, as is commonly thought.
... Climate change will have intricate and challenging effects on disease risk that will be difficult to predict, especially because of the intrinsic difficulty to predict the emergence of infectious diseases (Stone, 2008). It is likely that these disruption mechanisms will interact with each other and other disturbances, resulting in varying pathogen-host-environment interactions (Roger et al., 2016;Carlson et al., 2020;IPBES, 2020;Cohen et al., 2020). ...
Book
Full-text available
This situation analysis presents a thorough, evidence-based examination of the relationship between wildlife and zoonosis, wildlife and emerging human pathogens and associated diseases, their origins, drivers, and risk factors. There is considerable divergence of opinion around the subject both within and outside the biodiversity conservation community and given the ontological challenges and highly different perspectives, contradictory narrative is unsurprising. Context is all-important and to clarify this in the analysis, the evidence of human diseases coming from wildlife is compared to diseases emerging from domestic animals and humans themselves, to provide context and proportions of the relative risk. The report highlights key knowledge, and provides perspective on where research, policy, interventions, and capacity building are needed to reduce risks of zoonoses and emergent animal-origin human diseases globally.
... For example, the destruction of natural habitats tends to drive wildlife out of their original living space and into contact with humans, thus increasing the risk of animalto-human disease transmission (Roe et al., 2020;McNeely, 2021;Pelley, 2021). Furthermore, anthropogenic climate change could directly lead to deadlier future pandemics, as many diseases spread faster (Carlson et al., 2021) or expand their range and active season under higher temperatures (Curseu et al., 2010). ...
Article
Full-text available
As the world contends with the far-ranging impacts of the COVID-19 pandemic, ongoing environmental crises have, to some extent, been neglected during the pandemic. One reason behind this shift in priorities is the scarcity mindset triggered by the pandemic. Scarcity is the feeling of having less than what is necessary, and it causes people to prioritize immediate short-term needs over long-term ones. Scarcity experienced in the pandemic can reduce the willingness to engage in pro-environmental behavior, leading to environmental degradation that increases the chance of future pandemics. To protect pro-environmental behavior, we argue that it should not be viewed as value-laden and effortful, but rather reconceptualized as actions that address a multitude of human needs including pragmatic actions that conserve resources especially during scarcity. To bolster environmental protection, systematic changes are needed to make pro-environmental behavior better integrated into people's lives, communities, and cities, such that it is more accessible, less costly, and more resilient to future disturbances.
Article
Given the unprecedented rates of global warming, widespread shifts in species’ distributions are anticipated to play a key role for their survival. Yet, current conservation policies often allocate priority to native species and their typical habitats, thereby preserving historic conditions rather than preparing for future species distributions. Policy initiatives aimed at halting biodiversity loss, such as the EU 2030 Biodiversity Strategy, could provide an opportunity to proactively integrate species’ future distributions into conservation objectives on an international scale, for example, by encouraging management actions that allow for species migration along projected paths of dispersal. Acknowledging climate‐tracking species as a conceptually distinct phenomenon, we qualitatively analyzed to which extent the EU 2030 Biodiversity Strategy and the legal framework it is embedded in, such as the Draft Note on Designation Criteria (ENV.D.3/JC) and the Framework for Blue and Green Infrastructure (SWD/2019/193), integrated climate‐trackers into their conservation objectives. We found that the Biodiversity Strategy did not explicitly incorporate future patterns of species distributions into conservation planning but emphasized the maintenance of historical community compositions and species distributions. While the commitments in these legal documents to nature restoration will certainly be helpful for biodiversity conservation today and after 2030, it may miss the chance to be a more comprehensive conservation policy.
Preprint
Full-text available
Massive biological databases of species occurrences, or georeferenced locations where a species has been observed, are essential inputs for modeling present and future species distributions. Location accuracy is often assessed by determining whether the observation geocoordinates fall within the boundaries of the declared political divisions. This otherwise simple validation is complicated by the difficulty of matching political division names to the correct geospatial object. Spelling errors, abbreviations, alternative codes, and synonyms in multiple languages present daunting name disambiguation challenges. The inability to resolve political division names reduces usable data and analysis of erroneous observations can lead to flawed results. Here, we present the Geographic Name Resolution Service (GNRS), an application for the correction, standardization and indexing of world political division names. The GNRS resolves political division names against a reference database that combines names and codes from GeoNames with geospatial object identifiers from the Global Administrative Areas Database (GADM). In a trial resolution of political division names extracted from >270 million species occurrences, only 1.9%, representing just 6% of occurrences, matched exactly to GADM political divisions in their original form. The GNRS was able to resolve, completely or in part, 92% of the remaining 378,568 political division names, or 86% of the full biodiversity occurrence dataset. In an assessment of geocoordinate accuracy for >239 million species occurrences, resolution of political divisions by the GNRS enabled detection of an order of magnitude more errors and an order of magnitude more error-free occurrences. By providing a novel solution to a major data quality impediment, the GNRS liberates a tremendous amount of biodiversity data for quantitative biodiversity research. The GNRS runs as a web service and can be accessed via an API, an R package, and a web-based graphical user interface. Its modular architecture is easily integrated into existing data validation workflows.
Preprint
Full-text available
Massive biological databases of species occurrences, or georeferenced locations where a species has been observed, are essential inputs for modeling present and future species distributions. Location accuracy is often assessed by determining whether the observation geocoordinates fall within the boundaries of the declared political divisions. This otherwise simple validation is complicated by the difficulty of matching political division names to the correct geospatial object. Spelling errors, abbreviations, alternative codes, and synonyms in multiple languages present daunting name disambiguation challenges. The inability to resolve political division names reduces usable data and analysis of erroneous observations can lead to flawed results. Here, we present the Geographic Name Resolution Service (GNRS), an application for the correction, standardization and indexing of world political division names. The GNRS resolves political division names against a reference database that combines names and codes from GeoNames with geospatial object identifiers from the Global Administrative Areas Database (GADM). In a trial resolution of political division names extracted from >270 million species occurrences, only 1.9%, representing just 6% of occurrences, matched exactly to GADM political divisions in their original form. The GNRS was able to resolve, completely or in part, 92% of the remaining 378,568 political division names, or 86% of the full biodiversity occurrence dataset. In an assessment of geocoordinate accuracy for >239 million species occurrences, resolution of political divisions by the GNRS enabled detection of an order of magnitude more errors and an order of magnitude more error-free occurrences. By providing a novel solution to a major data quality impediment, the GNRS liberates a tremendous amount of biodiversity data for quantitative biodiversity research. The GNRS runs as a web service and can be accessed via an API, an R package, and a web-based graphical user interface. Its modular architecture is easily integrated into existing data validation workflows.
Article
Exposure and susceptibility underlie every organism’s parasite burden, and an untold diversity of factors can drive variation in both. Often, both exposure and susceptibility change in response to a given factor, and they can interact, such that their relative contributions to observed disease dynamics are obscured. These independent and interlinked changes often complicate empirical inference in disease ecology and ecoimmunology. Although many disease ecology studies address this problem, it is often implicit rather than explicit and requires a specific set of tools to tackle. Moreover, as yet, there is no established conceptual framework for disentangling susceptibility and exposure processes. Here, we consolidate previous theory and empirical understanding regarding the entwined effects of susceptibility and exposure, which we refer to as “the Twin Pillar Problem”. We provide a framework for conceptualising exposure‐susceptibility interactions, where they obscure, confound, induce, or counteract one another, providing some well‐known examples for each complicating mechanism. We synthesise guidelines for anticipating and controlling for covariance between exposure and susceptibility, and we detail statistical and operational methodology that researchers have employed to deal with them. Finally, we discuss novel emerging frontiers in their study in ecology, and their potential for further integration in the fields of wildlife and human health.
Article
Full-text available
Outbreaks of zoonotic diseases are accelerating at an unprecedented rate in the current era of globalization, with substantial impacts on the global economy, public health, and sustainability. Alien species invasions have been hypothesized to be important to zoonotic diseases by introducing both existing and novel pathogens to invaded ranges. However, few studies have evaluated the generality of alien species facilitating zoonoses across multiple host and parasite taxa worldwide. Here, we simultaneously quantify the role of 795 established alien hosts on the 10,473 zoonosis events across the globe since the 14 th century. We observe an average of~5.9 zoonoses per alien zoonotic host. After accounting for species-, disease-, and geographic-level sampling biases, spatial autocorrelation, and the lack of independence of zoonosis events, we find that the number of zoonosis events increase with the richness of alien zoonotic hosts, both across space and through time. We also detect positive associations between the number of zoonosis events per unit space and climate change, land-use change, biodiversity loss, human population density, and PubMed citations. These findings suggest that alien host introductions have likely contributed to zoonosis emergences throughout recent history and that minimizing future zoonotic host species introductions could have global health benefits.
Article
Parasites that infect multiple species cause major health burdens globally, but for many, the full suite of susceptible hosts is unknown. Predicting undocumented host‐parasite associations will help expand knowledge of parasite host specificities, promote the development of theory in disease ecology and evolution, and support surveillance of multi‐host infectious diseases. Analysis of global species interaction networks allows for leveraging of information across taxa, but link prediction at this scale is often limited by extreme network sparsity and lack of comparable trait data across species. Here we use recently developed methods to predict missing links in global mammal parasite networks using readily available data: network properties and evolutionary relationships among hosts. We demonstrate how these link predictions can efficiently guide the collection of species interaction data and increase the completeness of global species interaction networks. We amalgamate a global mammal host‐parasite interaction network (>29,000 interactions) and apply a hierarchical Bayesian approach for link prediction that leverages in formation on network structure and scaled phylogenetic distances among hosts. We use these predictions to guide targeted literature searches of the most likely yet undocumented interactions, and identify empirical evidence supporting many of the top “missing” links. We find that link prediction in global host‐parasite networks can successfully predict parasites of humans, domesticated animals, and endangered wildlife, representing a combination of published interactions missing from existing global databases, and potential but currently undocumented associations. Our study provides further insight into the use of phylogenies for predicting host‐parasite interactions, and highlights the utility of iterated prediction and targeted search to efficiently guide the collection of information on host‐parasite interactions. These data are critical for understanding the evolution of host specificity, and may be used to support disease surveillance through a process of predicting missing links, and targeting research towards the most likely undocumented interactions.
Article
Full-text available
Effective public health research and preparedness requires an accurate understanding of which virus species possess or are at risk of developing human transmissibility. Unfortunately, our ability to identify these viruses is limited by gaps in disease surveillance and an incomplete understanding of the process of viral adaptation. By fitting boosted regression trees to data on 224 human viruses and their associated traits, we developed a model that predicts the human transmission ability of zoonotic viruses with over 84% accuracy. This model identifies several viruses that may have an undocumented capacity for transmission between humans. Viral traits that predicted human transmissibility included infection of nonhuman primates, the absence of a lipid envelope, and detection in the human nervous system and respiratory tract. This predictive model can be used to prioritize high-risk viruses for future research and surveillance, and could inform an integrated early warning system for emerging infectious diseases.
Article
Full-text available
Host shifts–where a pathogen jumps between different host species–are an important source of emerging infectious disease. With on-going climate change there is an increasing need to understand the effect changes in temperature may have on emerging infectious disease. We investigated whether species’ susceptibilities change with temperature and ask if susceptibility is greatest at different temperatures in different species. We infected 45 species of Drosophilidae with an RNA virus and measured how viral load changes with temperature. We found the host phylogeny explained a large proportion of the variation in viral load at each temperature, with strong phylogenetic correlations between viral loads across temperature. The variance in viral load increased with temperature, while the mean viral load did not. This suggests that as temperature increases the most susceptible species become more susceptible, and the least susceptible less so. We found no significant relationship between a species’ susceptibility across temperatures, and proxies for thermal optima (critical thermal maximum and minimum or basal metabolic rate). These results suggest that whilst the rank order of species susceptibilities may remain the same with changes in temperature, some species may become more susceptible to a novel pathogen, and others less so.
Article
Full-text available
Background Usutu virus (USUV) is a mosquito-borne flavivirus, reported in many countries of Africa and Europe, with an increasing spatial distribution and host range. Recent outbreaks leading to regional declines of European common blackbird (Turdus merula) populations and a rising number of human cases emphasize the need for increased awareness and spatial risk assessment. Methods Modelling approaches in ecology and epidemiology differ substantially in their algorithms, potentially resulting in diverging model outputs. Therefore, we implemented a parallel approach incorporating two commonly applied modelling techniques: (1) Maxent, a correlation-based environmental niche model and (2) a mechanistic epidemiological susceptible-exposed-infected-removed (SEIR) model. Across Europe, surveillance data of USUV-positive birds from 2003 to 2016 was acquired to train the environmental niche model and to serve as test cases for the SEIR model. The SEIR model is mainly driven by daily mean temperature and calculates the basic reproduction number R0. The environmental niche model was run with long-term bio-climatic variables derived from the same source in order to estimate climatic suitability. Results Large areas across Europe are currently suitable for USUV transmission. Both models show patterns of high risk for USUV in parts of France, in the Pannonian Basin as well as northern Italy. The environmental niche model depicts the current situation better, but with USUV still being in an invasive stage there is a chance for under-estimation of risk. Areas where transmission occurred are mostly predicted correctly by the SEIR model, but it mostly fails to resolve the temporal dynamics of USUV events. High R0 values predicted by the SEIR model in areas without evidence for real-life transmission suggest that it may tend towards over-estimation of risk. Conclusions The results from our parallel-model approach highlight that relying on a single model for assessing vector-borne disease risk may lead to incomplete conclusions. Utilizing different modelling approaches is thus crucial for risk-assessment of under-studied emerging pathogens like USUV. Electronic supplementary material The online version of this article (10.1186/s12942-018-0155-7) contains supplementary material, which is available to authorized users.
Article
Full-text available
Methods for modeling species’ distributions in nature are typically evaluated empirically with respect to data from observations of species occurrence and, occasionally, absence at surveyed locations. Such models are relatively “theory‐free.” In contrast, theories for explaining species’ distributions draw on concepts like fitness, niche, and environmental suitability. This paper proposes that environmental suitability be defined as the conditional probability of occurrence of a species given the state of the environment at a location. Any quantity that is proportional to this probability is a measure of relative suitability and the support of this probability is the niche. This formulation suggests new methods for presence‐background modeling of species distributions that unify statistical methodology with the conceptual framework of niche theory. One method, the plug‐and‐play approach, is introduced for the first time. Variations on the plug‐and‐play approach were studied with respect to their numerical performance on 106 species from an exhaustively sampled presence–absence survey of vegetation in the Canton of Vaud, Switzerland. Additionally, we looked at the robustness of these methods to the presence of irrelevant information and sample size. Although irrelevant variables eroded the predictive performance of all methods, these methods were found to be both numerically and statistically robust.
Article
Full-text available
As the outbreak of Ebola virus disease (EVD) in West Africa is now contained, attention is turning from control to future outbreak prediction and prevention. Building on a previously published zoonotic niche map (Pigott et al., 2014), this study incorporates new human and animal occurrence data and expands upon the way in which potential bat EVD reservoir species are incorporated. This update demonstrates the potential for incorporating and updating data used to generate the predicted suitability map. A new data portal for sharing such maps is discussed. This output represents the most up-to-date estimate of the extent of EVD zoonotic risk in Africa. These maps can assist in strengthening surveillance and response capacity to contain viral haemorrhagic fevers.
Article
Full-text available
We created a new dataset of spatially interpolated monthly climate data for global land areas at a very high spatial resolution (approximately 1 km 2). We included monthly temperature (minimum, maximum and average), precipitation, solar radiation, vapour pressure and wind speed, aggregated across a target temporal range of 1970–2000, using data from between 9000 and 60 000 weather stations. Weather station data were interpolated using thin-plate splines with covariates including elevation, distance to the coast and three satellite-derived covariates: maximum and minimum land surface temperature as well as cloud cover, obtained with the MODIS satellite platform. Interpolation was done for 23 regions of varying size depending on station density. Satellite data improved prediction accuracy for temperature variables 5–15% (0.07–0.17 ∘ C), particularly for areas with a low station density, although prediction error remained high in such regions for all climate variables. Contributions of satellite covariates were mostly negligible for the other variables, although their importance varied by region. In contrast to the common approach to use a single model formulation for the entire world, we constructed the final product by selecting the best performing model for each region and variable. Global cross-validation correlations were ≥ 0.99 for temperature and humidity, 0.86 for precipitation and 0.76 for wind speed. The fact that most of our climate surface estimates were only marginally improved by use of satellite covariates highlights the importance having a dense, high-quality network of climate station data.
Article
Full-text available
Distributions of Earth’s species are changing at accelerating rates, increasingly driven by human-mediated climate change. Such changes are already altering the composition of ecological communities, but beyond conservation of natural systems, how and why does this matter? We review evidence that climate-driven species redistribution at regional to global scales affects ecosystem functioning, human well-being, and the dynamics of climate change itself. Production of natural resources required for food security, patterns of disease transmission, and processes of carbon sequestration are all altered by changes in species distribution. Consideration of these effects of biodiversity redistribution is critical yet lacking in most mitigation and adaptation strategies, including the United Nation’s Sustainable Development Goals.
Chapter
Full-text available
Here we examine the question of whether bats are truly unique as viral reservoirs as compared to other mammalian hosts. First, we examine what criteria are typically used to determine whether a given group of hosts is “special” or “unique” as disease reservoir hosts. Second, we address factors that may confound such determination, such as sampling or research bias. Third, we give a brief overview of the large body of published work that has focused on viral discovery in bats over the last 7 years. Fourth, we examine the total viral richness and proportion of zoonotic viruses that bats harbour in relation to all mammals. Fifth, we examine and review the literature published on the ecological, behavioral, and life history traits of bats and their relationship to viral diversity, and discuss this in the context of life-history traits of other mammals. Sixth, we summarize modeling efforts to better understand global patterns of viral richness, from a host species and spatial perspective, in bats and other mammals. Lastly, we summarize our findings, and discuss avenues for future research to more systematically understand wildlife disease reservoirs and the role of bats.
Article
Full-text available
The ecological niche is the set of environments in which a population of a species can persist without introduction of individuals from other locations. A good mathematical or computational representation of the niche is a prerequisite to addressing many questions in ecology, biogeography, evolutionary biology and conservation. A particularly challenging question for ecological niche modelling is the problem of presence-only modelling. That is, can an ecological niche be identified from records drawn only from the set of niche environments without records from non-niche environments for comparison? Here, I introduce a new method for ecological niche modelling from presence-only data called range bagging. Range bagging draws on the concept of a species' environmental range, but was inspired by the empirical performance of ensemble learning algorithms in other areas of ecological research. This paper extends the concept of environmental range to multiple dimensions and shows that range bagging is computationally feasible even when the number of environmental dimensions is large. The target of the range bagging base learner is an environmental tolerance of the species in a projection of its niche and is therefore an ecologically interpretable property of a species' biological requirements. The computational complexity of range bagging is linear in the number of examples, which compares favourably with the main alternative, Qhull. In conclusion, range bagging appears to be a reasonable choice for niche modelling in applications in which a presence-only method is desired and may provide a solution to problems in other disciplines where one-class classification is required, such as outlier detection and concept learning.
Article
The unexplained decline of amphibian populations across the world was first recognised in the late 20th century. When investigated, most of these “enigmatic” declines have been shown to be due to one of two types of infectious disease: ranavirosis caused by infection with FV3-like ranavirus or with common midwife toad virus, or chytridiomycosis caused by infection with Batrachochytrium dendrobatidis or B. salamandrivorans. In all cases examined, infection has been via the human-mediated introduction of the pathogen to a species or population in which it has not naturally co-evolved. While ranaviruses and B. salamandrivorans have caused regionally localised amphibian population declines in Europe, the chytrid fungus, B. dendrobatidis, has caused catastrophic multi-species amphibian population declines and species extinctions globally. These diseases have already caused the loss of amphibian biodiversity, and over 40% of known amphibian species are threatened with extinction. If this biodiversity loss is to be halted, it is imperative that regulations are put in place – and enforced – to prevent the spread of known and yet-to-be discovered amphibian pathogens. Also, it is incumbent on those who keep or study amphibians to take measures to minimise the risk of disease spread, including from captive animals to those in the wild.